Certain types of road crossings reduce longitudinal movement of stream fishes. We characterized the genetic structure of warm water fish populations as a potential indication of movement barriers in upland streams of the Ouachita National Forest. Fin-clip tissue samples of longear sunfish, Lepomis megalotis and highland stoneroller, Campostoma spadiceum were collected for microsatellite analysis. Initially, we selected two streams with road crossings and two without for reference; however, we added another reference stream because a natural waterfall on one of the reference streams resulted in a measurable barrier-effect. For each stream, a total of 75 to 90 fish of each species representing upstream, downstream, and mid-stream reaches were collected (a total of approximately 800 specimens). DNA was extracted and seven microsatellite primers were selected for highland stonerollers and ten for longear sunfish because the latter showed fewer alleles per loci (Na=11.7 versus 6.2, respectively). We detected an apparent disruption to gene flow in at least three of the five streams (although certain analyses also detected a disruption in a fourth stream and analyses are not yet complete for stonerollers) . Bayesian analysis of the longear sunfish samples predicted a maximum ΔK value at K = 10 indicating 10 distinguishable populations among the 15 sampled sites. This finding was consistent with an FSTAT pair-wise population assignment indicating three longear sunfish populations in Bear Creek, two in Crystal Prong, two in Long Creek, and two in Little Missouri; whereas, Blaylock appeared to have a single population. Populations from all three sites within Bear were distinguishable, including two sites separated by a piped-culvert. A barrier-effect was also detected in Crystal Prong which is a reference stream that has dry reaches, like Bear Creek, in summer. A natural waterfall on the Little Missouri reference stream also appeared as a barrier to gene flow with the two sites upstream of the falls representing a single population that differed significantly from the downstream population. Two populations were detected in Long Creek with the middle reach similar to both the upstream and the downstream, but the upstream differed significantly from the downstream population. Interestingly, the downstream road crossings in Long Creek were box culverts designed to facilitate fish passage; whereas, upstream crossings included the piped-culvert design known to reduce passage of warm water fishes. We detected no barriers to gene flow for Blaylock Creek, a reference stream that flows all summer like Long Creek and Little Missouri. We conclude that improved understanding of genetic structure in warm water fish populations can contribute to our assessment of movement barriers by showing disruption to natural patterns of gene flow in headwater upland streams.